Removing soldered CPU and using L2/PDS upgrade

[Franklinstein]

Has anybody ever tried removing the onboard CPU in a machine like the 6100 or 6400 and installing an L2 or PDS CPU upgrade card? Would one or both of these types of CPU upgrade work? Theoretically they both attach directly to the 60x bus and somehow completely take over the bus and/or disable the onboard CPU, which should mean the upgrade card has full access to ROM and other devices even at initial boot. 

Mostly I'm wondering if it would be possible to take the processor out of, say, a 6100/60 to safely be able to crank the original 30MHz bus to 40MHz without worrying about the original CPU being overclocked and consuming power/producing heat when it isn't doing anything useful. I have an 8100/110 board with a damaged CPU that I could use to test this (and up its 37MHz bus to 40MHz while I'm at it) but I'm curious if someone else has already been there and done that. The only concern with my 8100 board is that it has a warning that the Peltier junction needs to be connected or the board may be damaged, but it doesn't specify if this would be electrical damage from the lack of a load on that circuit, or if it's because of potential overheating damage to the processor, which won't be an issue since it won't be installed. 

 

[Franklinstein]

I decided to try with this 8100 board. It had both a cracked CPU and had suffered from a leaked battery, so nothing was lost from experimenting on it.

I removed the CPU (it came off in pieces since its ceramic substrate was shattered) and cleaned the board as well as I could. Some of the RAM SIMM slots are a little ugly but I was only testing with onboard RAM so I don't think there was much effect from tarnished contacts on unused slots. The board powered up without a CPU or an upgrade card, so I figured that was good. Then I tried booting with a 250/512k Crescendo card, and the board powered up but nothing happened: speaker made a little click when the reset button was pressed and a couple chips on the logic board got warm, but the Crescendo card's heat sink stayed pretty much cold. I tried with a beige G3's ROM module and it didn't do anything either. I looked at the 6100 Dev Note and the block diagram suggests that perhaps the HMC is only accessed directly by the CPU with no connection available through the PDS. I'm not sure if that's entirely accurate or not, since everything should have access to the same A and D pins as available on the 60x bus, but maybe there are special signal pins that were missing. Also it's possible that the CPU needs to perform initial configuration before the Crescendo card can perform its hostile takeover: there are a few pins on the PDS slot pinout that suggest the Crescendo card may have to be interrogated before it can assert control over the bus and/or some pins need to be set high or low before the Crescendo will come to life. If I had more info regarding exactly how the PDS accelerators functioned, maybe I could get it to work by bridging a few CPU pads or something. I dunno.

It's possible that the L2-style accelerators could work without a host CPU since they would have direct access to both the 60x bus and the memory controller if indeed the memory controller isn't entirely available from the 60x bus. I don't have any damaged Gazelle, Alchemy, or Tanzania boards to experiment with so that's a project for another time.